Logarithmic approximate multipliers

Množenje je ena izmed najpogosteje uporabljenih aritmetičnih operacij. Množilniki imajo pomemben vpliv na hitrost in porabo moči aritmetičnih vezij. V mnogih aplikacijah pa natančni rezultati niso vedno potrebni. Algoritmi in aplikacije, odporni na napake, so spodbudili razvoj približnih množilnikov, ker na ta način zmanjšamo porabo energije. Diplomsko delo se osredotoča na implementacijo in primerjavo približnih množilnikov v logaritemskem številskem sistemu. Izvedene so implementacije naslednjih množilnikov: Mitchellov algoritem (MA), Mitchellov algoritem z razstavljanjem operandov (OD-MA), množilnik z 2-območnim algoritmom (2-region) in množilnik s 3-območnim algoritmom (3-region). Glede na kriterija napake aproksimacije in kompleksnosti implementacije je narejena primerjava implementacij množilnikov.Multiplication is one of the most commonly used arithmetic operations. Multipliers have a significant impact on the speed and power dissipation of arithmetic circuits. However, in many applications precise results are not always required. Error-tolerant algorithms and applications encouraged the development of approximate multipliers because in this way the resource requirements are reduced. The thesis focuses on the implementation and comparison of some basic approximate multipliers in the logarithmic number system. The following multipliers are implemented: Mitchell\u27s algorithm based multiplier (MA), Operand decomposition Mitchell-based algorithm (OD-MA) multiplier, 2-region algorithm multiplier (2-region) and 3-region algorithm multiplier (3-region). According to the criteria of approximation error and implementation complexity, a comparison of implementations of multipliers is made